ControlWar: The Software War After Softwar and No Software

Oracle fought for the system of record. Salesforce fought for the system of engagement. The AI era is opening the fight for governed action.

That thread connects Larry Ellison, Marc Benioff, and the platform fight taking shape today around Bill McDermott, ServiceNow, Microsoft, Salesforce, SAP, Oracle, and the frontier model companies. I call this new phase ControlWar.

ControlWar is not only a ServiceNow story. ServiceNow is one of the clearest examples because McDermott has put unusually sharp language around the moment, talking about AI chaos, governed autonomous work, and a control tower that coordinates agents across the enterprise. The larger story runs past any one company’s product narrative. Enterprise software is moving toward a new control layer, and the prize this time is governed action.

Softwar, the war for the system of record

Larry Ellison understood enterprise software as a power game earlier than most. Oracle was never only a database company in the narrow technical sense. The database was where corporate truth lived, and if it became the foundation for applications, reporting, transactions, and decisions, then whoever owned it held a critical point of leverage.

That is why Softwar still matters as an artifact. Matthew Symonds wrote it in 2003 with running commentary from Ellison himself, and it reads less like a biography than a portrait of enterprise software as combat (Symonds & Ellison, 2003). Ellison did not treat categories as neutral markets. He treated them as contested terrain, and the database was the original enterprise control point.

In the 1990s and early 2000s, enterprise software had weight. It was installed, customized, and parked inside the data center. Buying it was not a transaction so much as a long commitment of infrastructure, consultants, integration work, and organizational patience. Oracle’s strength came from that permanence. Once a company tied its data and processes to Oracle, the relationship was hard to unwind, and the database stopped being a technical layer and became institutional gravity.

The PeopleSoft fight made that visible. Oracle spent close to eighteen months and roughly ten billion dollars to take a rival by force, and the battle revealed that enterprise software had matured into a consolidation game (Symonds & Ellison, 2003). The strategic question was no longer whether a company needed software. It was who would own the durable layers underneath the company. Oracle’s answer was to control the data, the applications, the customer relationship, and the economics. That was Softwar, the war for the system of record.

No Software, the war for the system of engagement

Marc Benioff came out of Oracle, and he did not try to beat Oracle by building a better on-premise Oracle. He attacked the premise instead. The “No Software” slogan worked because it compressed a technical shift, a business-model shift, and a cultural rebellion into two words. It never meant the literal end of software. It meant the end of software as customers had come to suffer it: the giant installation, the painful upgrade, the shelfware nobody used, the multi-year rollout, and the dependence on IT before a business user could see a single result.

The frustration had been building for years, and Salesforce gave it a name. Behind the slogan sat a different architecture. Software as a service moved the product from the customer’s data center to the vendor’s cloud, turned deployment from a project into a subscription, made upgrades a routine rhythm rather than a traumatic event, and let business teams prove value before they carried the full weight of installed software. Benioff was not only selling CRM in the cloud. He was selling a different relationship between the company and its software.

Oracle’s world handed control to IT, finance, and the infrastructure owners. Salesforce handed more of it to the business user, and the front office became programmable. Oracle had fought for the system of record. Salesforce fought for the system of engagement, and for a while software as a service felt like liberation.

The SaaS paradox

Every successful rebellion builds its own orthodoxy. Software as a service began as the cure for enterprise complexity and slowly produced a new kind of it. Each department bought its own cloud, each function adopted its own workflow, and each application brought its own data model, permissions, automation logic, dashboards, and reporting. The company escaped one monolith and ended up living inside a hundred smaller ones.

For years this stayed manageable because people carried the context between systems. Employees retyped information from one application into another, managers reconciled the reports that did not agree, analysts rebuilt the dashboards, and the long-tenured staff remembered which field could not be trusted and which customer had unusual terms. The enterprise ran because humans quietly absorbed the fragmentation.

The numbers behind that fragmentation are not small. Salesforce’s own 2026 connectivity research put the average enterprise at 957 applications with only 27% of them connected, and found companies running about a dozen AI agents each, half of them operating in isolation rather than as coordinated systems (Salesforce, 2026). Those figures come from a vendor selling the cure, so read them as motive-aware, but the direction is hard to dispute.

A human clicking between tabs is merely inefficient. An AI agent moving across those same disconnected systems without clear authority is a different category of thing. It becomes a governance question, a security question, a compliance question, and an accountability question all at once. The moment AI begins to act, fragmentation stops being a productivity drag and starts being a liability. That is the opening for ControlWar.

ControlWar, the war for governed action

The next control point is not the database alone, and it is not the SaaS application alone. It is the governed action layer that runs across systems, the place where AI agents, identity, memory, policy, permissions, audit, and execution all have to meet. The question underneath it is short. When AI acts inside a company, who governs the action?

That question is larger than model quality. Frontier models will keep getting better, cheaper, and faster, and enterprises will still care which ones they use. Model quality alone will not settle the enterprise market, because the hard problem is not whether AI can produce an answer. The hard problem is whether AI can safely complete real work inside a real company. A wrong answer is an inconvenience. A wrong action is a loss.

The data already says as much. In McKinsey’s 2025 State of AI survey, 51% of organizations using AI reported at least one negative consequence in the past year, most often inaccuracy, and the danger compounds the instant inaccuracy is allowed to act rather than only advise (McKinsey & Company, 2025).

Completing work safely takes more than intelligence. It takes context about the business, authority to act, a verified identity for the actor, awareness of the workflow, memory of what came before, an audit trail of what happened, a way to handle exceptions, and a way to roll back when something goes wrong. The enterprise does not buy intelligence in the abstract. It buys accountable outcomes. That is why the action layer is the prize.

There is a sharper version of this point that most vendor decks skip. Removing friction is not always a gain. I argued in January that institutions learn through friction, because the bottleneck a junior analyst hits and the contract clause an associate struggles with are the signals that tell an organization where it needs to redesign itself (Reddy, 2026a).

When AI absorbs all of that execution, throughput goes up and signal visibility goes down, and a company can hyper-optimize itself around assumptions that no longer hold. That is what I called frictionless fragility. The governed action layer matters precisely because it is where a company decides which friction to keep, which signals to preserve, and which actions need a human in the loop before they execute.

ServiceNow as the sharpest current example

Bill McDermott’s ServiceNow is one of the most important companies to watch in this phase, and not because it claims AI will make employees more productive. The stronger claim is that enterprises need a platform for governed autonomous work, which is a control-plane argument. McDermott put it bluntly at the company’s 2026 customer event: governance is not a feature, it is the whole ballgame, because without it the whole company can come down (BizTech Magazine, 2026).

ServiceNow spent years positioning itself as a workflow fabric across IT, HR, customer service, security, risk, and procurement. Before AI that made it a platform for digitizing work. In the AI era McDermott is trying to lift that into something larger, the operating layer for autonomous work, and the language is deliberate. The “AI Control Tower” is pitched as a single command center to govern, secure, and measure any AI agent, model, or workflow, including agents that did not come from ServiceNow (Computer Weekly, 2025; FinancialContent, 2026).

McDermott has gone as far as calling ServiceNow the “AI of the AIs” and selling a guarantee on AI go-lives in under 100 days, while pointing investors at a thirty-billion-dollar-plus subscription target for 2030 with AI expected to reach roughly 30% of contract value (Yahoo Finance, 2026). That last figure is a company projection aimed at investors, so treat it as ambition rather than fact, but the ambition itself is the signal.

There is a useful irony in who is making the argument. McDermott ran SAP before this, and he is now using a workflow platform to claim the layer above the systems of record he used to sell. His line for it is that while AI thinks, workflow acts (No Jitter, 2026). That does not mean ServiceNow wins. It means ServiceNow has named the central problem clearly: AI without a governed action layer becomes enterprise chaos. The terrain around that claim is crowded.

A crowded battlefield

Each contender enters the action layer from a different asset, and each makes a different control claim.

The detail matters, so it is worth walking through.

Microsoft is fighting from identity and the place where work already happens. Its Agent 365 is pitched in plain terms as the control plane for agents, generally available on May 1, 2026 at fifteen dollars per user per month, and bundled into a new Frontier Suite alongside Copilot, the Entra identity suite, and a context layer called Work IQ (Microsoft, 2026a; Quisitive, 2026).

The sharp end is identity. Microsoft now issues each agent its own Entra identity so that permissions, conditional access, and data-loss policies attach to the agent the way they attach to an employee (Microsoft, 2026b). The pitch lands because of what it exposes. One Fortune 500 manufacturer, after turning on the new agent scope policies, discovered that three of its forty-seven agents held permission to delete files that no one had intended to grant (Windows News, 2026). If work begins inside Microsoft 365, Microsoft has a natural claim on governing it.

Salesforce fights from the customer relationship and the data around it. Agentforce reached roughly eight hundred million dollars in annual recurring revenue by the close of its 2026 fiscal year, up 169% year over year, and Benioff has started describing Salesforce as the operating system for the agentic enterprise (Salesforce, 2026). Its control claim runs through trust. Salesforce routes every agent action through a trust layer that checks policy, grounding, security, and privacy before the action completes, which is the kind of plumbing that let a company like Williams-Sonoma deploy agents across its brands without losing its nerve (VentureBeat, 2025). If the agentic enterprise organizes around revenue and service, Salesforce stays central.

SAP comes at this from the transactional core. At Sapphire 2026 it framed its whole direction as the Autonomous Enterprise and turned Joule from a copilot into an agentic platform with more than forty specialized agents and 2,400 skills, embedded directly in the ERP that books the company’s money (SAP, 2026; SAVIC Technologies, 2026). The honest part of SAP’s story is also the revealing one. Most Joule agents still require a human to confirm before they execute a consequential action (ERP Pilot, 2026). If the most consequential actions still land in the system of record, SAP cannot be ignored.

Oracle holds the database and the cloud beneath it. At its AI World Tour in April 2026 it shipped twelve Fusion agentic applications for finance and supply chain that decide and execute inside guardrails rather than merely recommend, escalating the complex cases to a human, and it gives its agent studio away with Fusion subscriptions to keep customers from looking elsewhere (Oracle, 2026). Ellison’s original instinct still applies, because proprietary enterprise data becomes more valuable, not less, once agents need trustworthy context to act on.

Google enters from the model and the cloud beneath it. Its Gemini Enterprise platform, rebuilt out of Vertex AI, packages the model, an agent builder, and governance into one managed layer, and Google is one of the originators of the Agent-to-Agent protocol that lets agents from different vendors talk to each other. The reach is real, because the model and the cloud arrive together, though Google carries less of the system-of-record gravity that anchors SAP and Oracle.

There are also two contenders most market maps leave out, and they come at the action layer from the execution side rather than the system-of-record side. Automation Anywhere laid out a four-layer autonomy stack at its Imagine ‘26 event: secure and manage at the bottom, reason and think above it, orchestrate and execute above that, and build and interact on top (Reddy, 2026b). UiPath comes from the same heritage and now positions its Maestro layer as a cross-vendor control plane for robots, agents, and people, and Gartner’s agentic-automation Magic Quadrant names Microsoft, ServiceNow, and UiPath as its three Leaders, which seats a supposed legacy vendor at the same table as the giants.

The interesting part is the heritage these two share. Real enterprise work does not live in clean, API-first systems. It crosses old screens, PDFs, SAP transactions, Salesforce records, ServiceNow tickets, email attachments, and manual approvals, and that mess is exactly where RPA was born. The strategic claim is to run autonomous processes across that fragmented landscape and connect probabilistic agents to deterministic processes, with one audit record across every process.

The weakness is the mirror image of the strength. Broad execution reach is not the same as deep semantic understanding, and SAP will always know SAP finance, Salesforce will always know CRM context, and ServiceNow will always know the service desk better than a cross-system runner can. The bots they were built on are also brittle, and AI-native rivals that drive any interface directly are pressuring that fabric from below.

Workday has the workforce and the financials it already holds. The frontier model companies press from raw intelligence, developer adoption, and the agent frameworks and protocols forming around them. And a new generation of startups attacks the gaps the incumbents leave open, in agent observability, governance, security, memory, evaluation, simulation, and compliance, which is exactly the kind of contested whitespace where new categories are born.

ControlWar will not resolve cleanly. Enterprise software rarely crowns a single winner. Large companies federate, they layer, they integrate, and they tolerate overlap when standardizing would cost too much politically and operationally. The likeliest outcome is a contested mesh of control planes rather than one throne. The direction is still clear. The next strategic layer is governed action.

Picks, shovels, and the layer above both

It helps to place these combatants on a value stack rather than a single battle line. In October 2025 I borrowed Geoffrey Moore’s framing and split the agentic economy into picks and shovels (Reddy, 2025c). The picks are the horizontal model and silicon providers selling raw cognitive energy, and they live with the long-run risk of commoditization. The shovels are the application vendors who embed AI into specific business workflows, and they hold a data-gravity advantage because the agent that lives inside a company’s own system of record understands that company’s context better than any outside model can.

ControlWar is what happens one layer up from that fight. Above the utility layer and the application layer sits the outcome the customer actually wants to buy, the finished business result rather than the tool. Selling that result safely is impossible without a governed action layer underneath it, because no one will pay for a guaranteed outcome that cannot be audited, bounded, or reversed. The pricing is already moving in that direction, from per-seat licensing toward charging for work the agent completes. Governed action is the precondition for the outcome economy, not a feature bolted onto it.

The organization on the other side of the war

There is a reason the control layer matters beyond risk management. The shape of the firm is changing underneath it. In November 2025 I argued that by 2030 the defining feature of a high-performance organization will not be headcount but leverage, because AI is collapsing coordination costs and compressing the mid-management layers that existed to move information up and down a hierarchy (Reddy, 2025d). As that mechanical oversight thins out, the individual contributor starts operating as an outcome node supported by a cluster of agents, and the scarce skill shifts from running people to designing the system they and their agents work inside.

That is the deeper stake in ControlWar. A flatter, more leveraged firm has fewer humans standing between intent and execution, which means the governance has to live in the architecture rather than in a chain of managers reviewing each other’s work. When hierarchy was thick, the hierarchy was the control plane. When hierarchy thins, something has to provide coherence in its place, and that something is the governed action layer. Leverage without architecture is just instability with better margins.

This also names a new human role. I have been describing it as the AI-Native Architect, the person who designs where intelligence lives, what an agent is allowed to do, where humans stay accountable, and how the system learns without losing trust (Reddy, 2026c). The same piece reframed the old software development lifecycle as what we now call the Adaptive Development Lifecycle, where signal, shaping, building, proving, and governing collapse into one loop because agents now participate in all of them. The control layer is what that architect actually builds. The market is fighting over the platform; the enterprise still needs someone inside it who can turn a control plane into a coherent operating model.

The case for skepticism, and why it strengthens the thesis

The strongest argument against everything above is that the agentic enterprise is mostly marketing, and that argument deserves a fair hearing because the evidence for it is real. Gartner has estimated that of the thousands of vendors claiming agentic AI, only around 130 are genuinely delivering autonomous systems, and it expects more than 40% of agentic AI projects to be cancelled by the end of 2027 on cost, unclear value, and weak risk controls (Gartner, 2025).

The research piles up from there. MIT’s NANDA initiative found that 95% of enterprise AI pilots delivered no measurable financial return (MIT NANDA, 2025). Carnegie Mellon’s benchmark found the best agents completing only about a third of real office tasks (Carnegie Mellon University, 2025). Forrester expects fewer than 15% of firms to switch on the agentic features already sitting in their automation suites, with ROI and governance worries keeping most on deterministic automation through 2026 (Forrester, 2026). Prompt injection remains unsolved enough that OpenAI has publicly acknowledged it may never be fully fixed.

Notice what every one of those failures is actually about. They are not failures of intelligence. They are failures of deployment, data quality, ownership of edge cases, and governance. The pilots that die in production die because no one built the layer that makes an agent’s work safe, legible, and reversible. The skeptic’s evidence is not a refutation of ControlWar. It is the bill of materials for it. A market wasting billions on ungoverned agents is a market telling you, in the loudest way available, where the missing layer is.

There is a subtler risk worth naming too, because it cuts against the instinct to deploy as fast as possible. If many companies run agents trained on similar data, pointed at similar objectives, inside the same competitive environment, those agents start to produce similar decisions. Harvard Business Review called this the agentic convergence trap, where the result of mass automation is not advantage but sameness, and in some markets something that looks like coordination without anyone coordinating (van Esch, Cui, & Black, 2026). The RealPage case is the cautionary example, where competing landlords each independently switched on an auto-accept feature and the aggregate result looked like coordinated pricing.

I have argued the corollary: proprietary data helps, but proprietary judgment matters more, and a pricing agent can lift margin while quietly eroding the very differentiation that justified the brand (Reddy, 2026d). Governance in the agentic era is therefore not only about safety. It is about keeping a company recognizably itself while its competitors optimize toward the same generic mean. That makes the action layer a strategic instrument, not just a compliance one.

Why this matters to 451

This is where 451’s interest begins. We are less interested in predicting which incumbent wins the next software war, because that question, while real, is not the most important one. The more important question is architectural. What does an AI-native enterprise actually need in order to let autonomous systems do real work without losing accountability, context, memory, and control?

In our view the enterprise AI conversation is still too centered on models. Models matter, but access to a powerful model does not make a company AI-native. A company becomes AI-native when it can turn intent into governed action, learn from the result, preserve institutional memory, and improve the system without breaking trust. That takes a real architecture: a way to capture intent, shape work before it executes, build and verify agentic workflows, prove what actually happened, and govern action across people, systems, and agents together.

Two further commitments shape how we read this fight, and both predate it in our own writing. The first is that the contest is not only commercial, it is about agency. In July 2025 I argued that AI is sorting the economy into intelligence owners, AI-amplified humans, and AI-subjected humans, and that the deepest inequality of this era is not income but agency, the gap between those who shape intelligent systems and those whose lives are shaped by systems they cannot inspect (Reddy, 2025b).

A control plane decides which side of that line an enterprise, and the people inside it, end up on. The same record that lets a board audit an agent is the record that lets an employee or a customer contest one. Governed action and accountable agency are the same architecture seen from two directions.

The second commitment is older still, going back to work we published in mid-2025: AI is not a product, it is a relationship, and the standard for it is regenerative rather than merely efficient (Reddy, 2025a). A system can get faster while getting more fragile, more productive while depleting the trust and judgment it runs on. The Regenerative Prosperity Model formalizes this with trust as the connective tissue between domains, proof as the discipline that verifies outcomes, and stewardship as the loop that returns value to the sources of value (Reddy, 2025e).

Read against ControlWar, those three are not soft ideas. Trust is what a control plane is for, proof is what an audit record produces, and stewardship is the difference between an enterprise that compounds capability and one that optimizes itself hollow. The leadership that can hold all of this together is what I have called Level 6, the capacity to govern living systems of human and machine intelligence without surrendering judgment to them (Reddy, 2026e). The control layer is the machinery. Regenerative intent is the reason to build it well.

The continuum

Three eras, one pattern.

Each era began by exposing the failure of the one before it. Database software made scattered information unbearable and created the opening for the system of record. Installed software made deployment too slow and too heavy and created the opening for the cloud. Now agentic AI is making fragmented, ungoverned work impossible to ignore, and it is creating the opening for the governed action layer.

Larry Ellison gave the industry Softwar. Marc Benioff gave it No Software. Bill McDermott is making one of the clearest incumbent claims in ControlWar. But ControlWar is bigger than ServiceNow. It is the fight over the operating model of the AI-native enterprise, and the prize is not another application category. It is the governed action layer where AI, people, workflows, systems, memory, and accountability finally have to meet. The companies and founders who win it will not be the ones that deploy the most agents. They will be the ones that can answer, for every agent they run, who acted and under what authority, and still keep the enterprise recognizably its own.

That is where the next software war will be fought. And that is where 451 intends to build.

---

References

BizTech Magazine. (2026, May 7). ServiceNow Knowledge 2026: Enterprise AI implementation needs guardrails and control. https://biztechmagazine.com/article/2026/05/servicenow-knowledge-2026-enterprise-ai-implementation-needs-guardrails-and-control

Carnegie Mellon University. (2025). TheAgentCompany: Benchmarking LLM agents on consequential real-world tasks.

https://the-agent-company.com

Computer Weekly. (2025, May 7). ServiceNow K25: McDermott vaunts agentic AI platform as revolutionary. https://www.computerweekly.com/news/366623680/ServiceNow-K25-McDermott-vaunts-agentic-AI-platform-as-revolutionary

ERP Pilot. (2026, March 1). AI in ERP systems 2026: What works? https://www.erp-pilot.com/erp/ai-in-erp

FinancialContent. (2026, March 31). The AI Control Tower: A deep dive into ServiceNow’s (NOW) GenAI evolution and 2026 outlook. https://markets.financialcontent.com/stocks/article/finterra-2026-3-31-the-ai-control-tower-a-deep-dive-into-servicenows-now-genai-evolution-and-2026-outlook

Forrester. (2026). Predictions 2026: Automation at the crossroads. https://www.forrester.com/blogs/predictions-2026-automation-at-the-crossroads

Gartner. (2025, June 25). Gartner predicts over 40% of agentic AI projects will be canceled by end of 2027 [Press release]. https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027

McKinsey & Company. (2025, November 5). The state of AI in 2025: Agents, innovation, and transformation. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai

Microsoft. (2026a, March 9). Powering frontier transformation with Copilot and agents. Microsoft 365 Blog. https://www.microsoft.com/en-us/microsoft-365/blog/2026/03/09/powering-frontier-transformation-with-copilot-and-agents/

Microsoft. (2026b). What’s new in Copilot Studio. Microsoft Learn. https://learn.microsoft.com/en-us/microsoft-copilot-studio/whats-new

MIT NANDA. (2025, July). The GenAI divide: State of AI in business 2025. MIT Media Lab, Project NANDA.

No Jitter. (2026, April 9). ServiceNow strides toward becoming the control plane for agentic business. https://www.nojitter.com/ai-automation/servicenow-strides-toward-becoming-the-control-plane-for-agentic-business

Oracle. (2026, April 9). Oracle introduces Fusion Agentic Applications for finance and supply chain. Oracle Newsroom. https://www.oracle.com/news/announcement/oracle-introduces-fusion-agentic-applications-for-finance-and-supply-chain-2026-04-09/

Quisitive. (2026, May 4). Microsoft 365 E7: The enterprise AI governance guide for CIOs and CISOs. https://quisitive.com/microsoft-365-e7-the-enterprise-ai-governance-guide-for-cios-and-cisos/

Reddy, S. (2025a, June 4). AI is not a product. It’s a relationship: Reimagining artificial intelligence through a regenerative lens. 451 Ventures. https://451ventures.substack.com/p/ai-is-not-a-product-its-a-relationship

Reddy, S. (2025b, July 26). The AI class structure: Inequality, agency, and power in the age of machine intelligence. 451 Ventures. https://451ventures.substack.com/p/the-ai-class-structure

Reddy, S. (2025c, October 8). The agentic AI revolution: An analysis of the coming market shift. 451 Ventures. https://451ventures.substack.com/p/the-agentic-ai-revolution

Reddy, S. (2025d, November 3). AI-native organization design 2030: Structural compression, human augmentation, and the redesign of the firm. 451 Ventures. https://451ventures.substack.com/p/ai-native-organization-design-2030

Reddy, S. (2025e, November 30). Regenerative Prosperity Model 2.0: From regenerative philosophy to regenerative operating system. 451 Ventures. https://451ventures.substack.com/p/regenerative-prosperity-model-20

Reddy, S. (2026a, January 9). Institutional evolution in the age of automated execution. 451 Ventures. https://451ventures.substack.com/p/institutional-evolution-in-the-age

Reddy, S. (2026b, May 19). Automation Anywhere and the architecture of the Autonomous Enterprise. 451 Ventures. https://451ventures.substack.com/p/automation-anywhere-and-the-architecture

Reddy, S. (2026c, March 24). The AI-native architect: A new human archetype for the age of agentic enterprise. 451 Ventures. https://451ventures.substack.com/p/the-ai-native-architect

Reddy, S. (2026d, May 14). Beyond agentic AI: The case for the governed autonomous enterprise. 451 Ventures. https://451ventures.substack.com/p/beyond-agentic-ai-the-case-for-the

Reddy, S. (2026e, May 29). Level 6 leadership: The leadership model for the age of agentic enterprise. 451 Ventures. https://451ventures.substack.com/p/level-6-leadership

SAP. (2026, May). SAP unveils the Autonomous Enterprise. SAP News Center. https://news.sap.com/2026/05/sap-sapphire-sap-unveils-autonomous-enterprise/

SAVIC Technologies. (2026, April 17). SAP Joule agentic AI 2026: Autonomous ERP agents explained. https://www.savictech.com/insights/sap-joule-agentic-ai-2026/

Salesforce. (2026, February 25). Salesforce delivers record fourth quarter fiscal 2026 results. Salesforce Investor Relations. https://investor.salesforce.com/news/news-details/2026/Salesforce-Delivers-Record-Fourth-Quarter-Fiscal-2026-Results/default.aspx

Symonds, M., & Ellison, L. (2003). Softwar: An intimate portrait of Larry Ellison and Oracle. Simon & Schuster.

van Esch, P., Cui, Y. G., & Black, J. S. (2026, May 13). Beware the agentic convergence trap. Harvard Business Review. https://hbr.org/2026/05/beware-the-agentic-convergence-trap

VentureBeat. (2025, December 23). While everyone talks about an AI bubble, Salesforce quietly added 6,000 enterprise customers in 3 months. https://venturebeat.com/technology/while-everyone-talks-about-an-ai-bubble-salesforce-quietly-added-6-000

Windows News. (2026). AI agents in 2026: Microsoft governance, identity perimeter, and email risk. https://windowsnews.ai/article/ai-agents-in-2026-microsoft-governance-identity-perimeter-and-email-risk.423028

Yahoo Finance. (2026, May 5). ServiceNow Investor Day: AI “Control Tower” pitch, $30B+ 2030 subscription target, margin upside. https://finance.yahoo.com/markets/stocks/articles/servicenow-investor-day-ai-control-030803021.html